[Eeglablist] averaging

[Arnaud Delorme]

Hi Tom,

your comments are correct.

>My concern is that for a given participant, there may be less artefact-
>free trials in Block 1 than Block 2 and this should be weighted 
>accordingly so that Block 2 contributes more to the individual average. 
>The grand average I would imagine to be a straightforward average of all 
>the individual averages in the group and should be not weighted. 
>  
>
The pop_comperp() function considers that you want to compute a grand 
average ERP across several subjects, not across two trial averages of 
the same subject. One of the reasons is that, in EEGLAB, we consider 
that all blocks of each subject should be concatenated (for running ICA 
for instance; note that you should not do that of course if a subject 
EEG has been recorded on 2 separate days). If you want to apply standard 
t-statistics across subjects, you cannot weights the trials average of 
each subject by the number of trials in each sub-average. Also, you do 
not want one subject with more trials to dominate the average. I 
acknowledge that ERP tools in EEGLAB are not optimal and that we should 
introduce the option you mentioned. Any other comments or suggestions 
are appreciated.

Best,

Arno
/
/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20050215/e0b68b59/attachment.html>